A new study published in Nature reveals that artificial intelligence systems are developing an unexpected ability to perform basic arithmetic, a task they were not explicitly trained for. Researchers from leading tech labs observed that large language models, when scaled to a sufficient size, spontaneously begin to solve elementary math problems with accuracy far beyond random chance. This emergent capability suggests that the models are forming internal representations of numerical concepts through pattern recognition in vast text datasets. The finding challenges previous assumptions about the limits of AI reasoning and opens new questions about what other latent skills might emerge as systems grow more complex. For the full details, read the complete article at https://technologyreview.com/2024/05/ai-emergent-arithmetic.